In [42]:
import numpy as np
from glob import glob
import itertools
from skimage.io import imread, imshow
from skimage.feature import canny
from skimage.filters import sobel, laplace, prewitt, roberts
from scipy import ndimage as ndi
import cv2
import matplotlib.pyplot as plt

Export Images

In [2]:
images_path = glob('images/*.jpg')
In [3]:
images_path
Out[3]:
['images/гриб.jpg',
 'images/images.jpg',
 'images/здание.jpg',
 'images/рентген.jpg']
In [6]:
images = list(map(imread, images_path))
In [7]:
fig, axs = plt.subplots(nrows=len(images), ncols=1, figsize=(8, 10))

for idx, image in enumerate(images):
    axs[idx].imshow(image);

Canny algorithm

In [12]:
def canny_segmentation(image, sigma=1, plot_edges=True, ax=None):
    gray = cv2.cvtColor(image, cv2.COLOR_BGR2GRAY)
    
    edges = canny(gray, sigma)
    
    if plot_edges:
        ax.imshow(edges);
    
    return edges

Default parameters

In [14]:
fig, axs = plt.subplots(nrows=len(images), ncols=1, figsize=(8, 10))

for idx, image in enumerate(images):
    canny_segmentation(image, ax=axs[idx])

Experiments with sigma

In [17]:
sigma_range = range(1, 10)
In [18]:
edges_images = [[canny_segmentation(img, sigma=sig, plot_edges=False) for sig in sigma_range]
               for img in images]
In [24]:
fig, axs = plt.subplots(nrows=len(images), ncols=len(sigma_range), figsize=(80, 40))

for i, j in itertools.product(range(len(images)), range(len(sigma_range))):
    axs[i][j].imshow(edges_images[i][j]);
    axs[i][j].set_title('Sigma: {}'.format(sigma_range[j]))

More detailed for each image

In [30]:
fig, axs = plt.subplots(nrows=len(sigma_range), ncols=1, figsize=(10, 40))

for idx, edge in enumerate(edges_images[0]):
    axs[idx].imshow(edge);
    axs[idx].set_title('Sigma: {}'.format(sigma_range[idx]))
In [31]:
fig, axs = plt.subplots(nrows=len(sigma_range), ncols=1, figsize=(10, 40))

for idx, edge in enumerate(edges_images[1]):
    axs[idx].imshow(edge);
    axs[idx].set_title('Sigma: {}'.format(sigma_range[idx]))
In [32]:
fig, axs = plt.subplots(nrows=len(sigma_range), ncols=1, figsize=(10, 40))

for idx, edge in enumerate(edges_images[2]):
    axs[idx].imshow(edge);
    axs[idx].set_title('Sigma: {}'.format(sigma_range[idx]))
In [34]:
fig, axs = plt.subplots(nrows=len(sigma_range), ncols=1, figsize=(10, 40))

for idx, edge in enumerate(edges_images[3]):
    axs[idx].imshow(edge);
    axs[idx].set_title('Sigma: {}'.format(sigma_range[idx]))

less sigma - more detailed edge detection

Sobel

In [38]:
def sobel_segmentation(image, plot_edges=True, ax=None):
    gray = cv2.cvtColor(image, cv2.COLOR_BGR2GRAY)
    
    elevation_map = sobel(gray)
    
    if plot_edges:
        ax.imshow(elevation_map);
    
    return elevation_map
In [40]:
sobel_edges_images = list(map(lambda x: sobel_segmentation(x, plot_edges=False), images))
In [41]:
fig, axs = plt.subplots(nrows=len(images), ncols=1, figsize=(8, 10))

for idx, image in enumerate(images):
    sobel_segmentation(image, ax=axs[idx])

Laplace operator

In [77]:
def laplace_segmentation(image, kernel_size=3, plot_edges=True, ax=None):
    edges = laplace(image, ksize=kernel_size)
    
    if plot_edges:
        ax.imshow(edges);
    
    return edges

Default parameters

In [78]:
fig, axs = plt.subplots(nrows=len(images), ncols=1, figsize=(8, 10))

for idx, image in enumerate(images):
    laplace_segmentation(image, ax=axs[idx])

Clipping input data to the valid range for imshow with RGB data ([0..1] for floats or [0..255] for integers).
Clipping input data to the valid range for imshow with RGB data ([0..1] for floats or [0..255] for integers).
Clipping input data to the valid range for imshow with RGB data ([0..1] for floats or [0..255] for integers).
Clipping input data to the valid range for imshow with RGB data ([0..1] for floats or [0..255] for integers).

Experiments with kernle size

In [79]:
ksize_range = range(3, 10)
In [80]:
laplace_edges_images = [[laplace_segmentation(img, kernel_size=kernel_size, plot_edges=False) 
                         for kernel_size in ksize_range]
                        for img in images]
In [81]:
fig, axs = plt.subplots(nrows=len(images), ncols=len(ksize_range), figsize=(80, 40))

for i, j in itertools.product(range(len(images)), range(len(ksize_range))):
    axs[i][j].imshow(laplace_edges_images[i][j]);
    axs[i][j].set_title('Kernel size: {}'.format(ksize_range[j]))

Clipping input data to the valid range for imshow with RGB data ([0..1] for floats or [0..255] for integers).
Clipping input data to the valid range for imshow with RGB data ([0..1] for floats or [0..255] for integers).
Clipping input data to the valid range for imshow with RGB data ([0..1] for floats or [0..255] for integers).
Clipping input data to the valid range for imshow with RGB data ([0..1] for floats or [0..255] for integers).
Clipping input data to the valid range for imshow with RGB data ([0..1] for floats or [0..255] for integers).
Clipping input data to the valid range for imshow with RGB data ([0..1] for floats or [0..255] for integers).
Clipping input data to the valid range for imshow with RGB data ([0..1] for floats or [0..255] for integers).
Clipping input data to the valid range for imshow with RGB data ([0..1] for floats or [0..255] for integers).
Clipping input data to the valid range for imshow with RGB data ([0..1] for floats or [0..255] for integers).
Clipping input data to the valid range for imshow with RGB data ([0..1] for floats or [0..255] for integers).
Clipping input data to the valid range for imshow with RGB data ([0..1] for floats or [0..255] for integers).
Clipping input data to the valid range for imshow with RGB data ([0..1] for floats or [0..255] for integers).
Clipping input data to the valid range for imshow with RGB data ([0..1] for floats or [0..255] for integers).
Clipping input data to the valid range for imshow with RGB data ([0..1] for floats or [0..255] for integers).
Clipping input data to the valid range for imshow with RGB data ([0..1] for floats or [0..255] for integers).
Clipping input data to the valid range for imshow with RGB data ([0..1] for floats or [0..255] for integers).
Clipping input data to the valid range for imshow with RGB data ([0..1] for floats or [0..255] for integers).
Clipping input data to the valid range for imshow with RGB data ([0..1] for floats or [0..255] for integers).
Clipping input data to the valid range for imshow with RGB data ([0..1] for floats or [0..255] for integers).
Clipping input data to the valid range for imshow with RGB data ([0..1] for floats or [0..255] for integers).
Clipping input data to the valid range for imshow with RGB data ([0..1] for floats or [0..255] for integers).
Clipping input data to the valid range for imshow with RGB data ([0..1] for floats or [0..255] for integers).
Clipping input data to the valid range for imshow with RGB data ([0..1] for floats or [0..255] for integers).
Clipping input data to the valid range for imshow with RGB data ([0..1] for floats or [0..255] for integers).
Clipping input data to the valid range for imshow with RGB data ([0..1] for floats or [0..255] for integers).
Clipping input data to the valid range for imshow with RGB data ([0..1] for floats or [0..255] for integers).
Clipping input data to the valid range for imshow with RGB data ([0..1] for floats or [0..255] for integers).
Clipping input data to the valid range for imshow with RGB data ([0..1] for floats or [0..255] for integers).

More detailed for each image

In [82]:
fig, axs = plt.subplots(nrows=len(ksize_range), ncols=1, figsize=(10, 40))

for idx, edge in enumerate(laplace_edges_images[0]):
    axs[idx].imshow(edge);
    axs[idx].set_title('Kernel size: {}'.format(ksize_range[idx]))

Clipping input data to the valid range for imshow with RGB data ([0..1] for floats or [0..255] for integers).
Clipping input data to the valid range for imshow with RGB data ([0..1] for floats or [0..255] for integers).
Clipping input data to the valid range for imshow with RGB data ([0..1] for floats or [0..255] for integers).
Clipping input data to the valid range for imshow with RGB data ([0..1] for floats or [0..255] for integers).
Clipping input data to the valid range for imshow with RGB data ([0..1] for floats or [0..255] for integers).
Clipping input data to the valid range for imshow with RGB data ([0..1] for floats or [0..255] for integers).
Clipping input data to the valid range for imshow with RGB data ([0..1] for floats or [0..255] for integers).
In [83]:
fig, axs = plt.subplots(nrows=len(ksize_range), ncols=1, figsize=(10, 40))

for idx, edge in enumerate(laplace_edges_images[1]):
    axs[idx].imshow(edge);
    axs[idx].set_title('Kernel size: {}'.format(ksize_range[idx]))

Clipping input data to the valid range for imshow with RGB data ([0..1] for floats or [0..255] for integers).
Clipping input data to the valid range for imshow with RGB data ([0..1] for floats or [0..255] for integers).
Clipping input data to the valid range for imshow with RGB data ([0..1] for floats or [0..255] for integers).
Clipping input data to the valid range for imshow with RGB data ([0..1] for floats or [0..255] for integers).
Clipping input data to the valid range for imshow with RGB data ([0..1] for floats or [0..255] for integers).
Clipping input data to the valid range for imshow with RGB data ([0..1] for floats or [0..255] for integers).
Clipping input data to the valid range for imshow with RGB data ([0..1] for floats or [0..255] for integers).
In [84]:
fig, axs = plt.subplots(nrows=len(ksize_range), ncols=1, figsize=(10, 40))

for idx, edge in enumerate(laplace_edges_images[2]):
    axs[idx].imshow(edge);
    axs[idx].set_title('Kernel size: {}'.format(ksize_range[idx]))

Clipping input data to the valid range for imshow with RGB data ([0..1] for floats or [0..255] for integers).
Clipping input data to the valid range for imshow with RGB data ([0..1] for floats or [0..255] for integers).
Clipping input data to the valid range for imshow with RGB data ([0..1] for floats or [0..255] for integers).
Clipping input data to the valid range for imshow with RGB data ([0..1] for floats or [0..255] for integers).
Clipping input data to the valid range for imshow with RGB data ([0..1] for floats or [0..255] for integers).
Clipping input data to the valid range for imshow with RGB data ([0..1] for floats or [0..255] for integers).
Clipping input data to the valid range for imshow with RGB data ([0..1] for floats or [0..255] for integers).
In [85]:
fig, axs = plt.subplots(nrows=len(ksize_range), ncols=1, figsize=(10, 40))

for idx, edge in enumerate(laplace_edges_images[3]):
    axs[idx].imshow(edge);
    axs[idx].set_title('Kernel size: {}'.format(ksize_range[idx]))

Clipping input data to the valid range for imshow with RGB data ([0..1] for floats or [0..255] for integers).
Clipping input data to the valid range for imshow with RGB data ([0..1] for floats or [0..255] for integers).
Clipping input data to the valid range for imshow with RGB data ([0..1] for floats or [0..255] for integers).
Clipping input data to the valid range for imshow with RGB data ([0..1] for floats or [0..255] for integers).
Clipping input data to the valid range for imshow with RGB data ([0..1] for floats or [0..255] for integers).
Clipping input data to the valid range for imshow with RGB data ([0..1] for floats or [0..255] for integers).
Clipping input data to the valid range for imshow with RGB data ([0..1] for floats or [0..255] for integers).

Prewitt segmentation

In [86]:
def prewitt_segmentation(image, plot_edges=True, ax=None):
    gray = cv2.cvtColor(image, cv2.COLOR_BGR2GRAY)
    
    edges = prewitt(gray)
    
    if plot_edges:
        ax.imshow(edges);
    
    return edges
In [87]:
prewitt_edges_images = list(map(lambda x: prewitt_segmentation(x, plot_edges=False), images))
In [88]:
fig, axs = plt.subplots(nrows=len(images), ncols=1, figsize=(8, 10))

for idx, image in enumerate(images):
    prewitt_segmentation(image, ax=axs[idx])

Roberts segmentation

In [89]:
def roberts_segmentation(image, plot_edges=True, ax=None):
    gray = cv2.cvtColor(image, cv2.COLOR_BGR2GRAY)
    
    edges = roberts(gray)
    
    if plot_edges:
        ax.imshow(edges);
    
    return edges
In [90]:
roberts_edges_images = list(map(lambda x: roberts_segmentation(x, plot_edges=False), images))
In [91]:
fig, axs = plt.subplots(nrows=len(images), ncols=1, figsize=(8, 10))

for idx, image in enumerate(images):
    roberts_segmentation(image, ax=axs[idx])
In [ ]:
In [ ]: